github →

New request

#733: Feature Request: Expose Hermes WebUI as a native MCP server to enable direct agent-to-WebUI communication and full headless control

We'll provision a sandbox, run an agent against the issue, and open a draft PR. You can pull the branch and iterate from there.

Issue
enhancementcli-paritygateway

Hermes Agent already has a native MCP server built in. This feature request proposes that the Hermes WebUI also exposes itself as an MCP server, which would allow Hermes Agent to communicate with the WebUI directly and natively, without any custom integration. This is the most natural architecture for the two to talk to each other and would make displaying live swarm data in the WebUI significantly simpler and more reliable.

Currently the WebUI has to poll or rely on WebSocket connections to get data from the agent layer. If the WebUI exposes an MCP server, Hermes Agent can push data to it directly using the same protocol it already uses for everything else. This removes the need for bespoke data pipelines between the agent and the UI.

What the WebUI MCP server should expose:

Every panel and action on the dashboard should have a corresponding MCP tool, making the WebUI fully headless and controllable via any MCP-compatible client including Claude Desktop:

  1. Agent status and management - query agent status, quality scores, hallucination rates, swap models, replace agents

  2. Cost and token tracking - query 7-day cost, monthly pace, budget remaining, cost per agent and per provider

  3. VPS health - query live CPU, RAM and disk usage

  4. Hallucination alerts - list active alerts, dismiss alerts, set and update thresholds per agent

  5. Team management - view, create and edit teams and agents, update roles, skills and model assignments, trigger config reload without restart

  6. MCP server management - enable, disable, add and remove MCP servers per agent

  7. Live event stream - query recent tool calls, errors, successes and warnings per agent

  8. Model health - compare models by quality score and hallucination rate, get replacement recommendations

  9. Aegis data - query traces, violations, pending approvals and anomalies from the Aegis gateway

  10. Heartbeat status - check whether the agent process is alive and when it was last seen

  11. Provider failover status - query which provider is currently serving each model and view failover history per agent

Why this is the right architecture:

Aegis already implements this pattern successfully, exposing an MCP server at ws://localhost:8080/mcp-audit with tools that map directly to its internal API. The Hermes WebUI could follow the same pattern. Since Hermes Agent speaks MCP natively, connecting it to a WebUI MCP server requires no new protocol work on the agent side at all. The two components would communicate as peers in the same ecosystem.

This would also allow any other MCP-compatible client such as Claude Desktop to query and control the entire swarm via natural language without opening the browser.

Reference: Aegis MCP server implementation shows a working example of this pattern.

Related: #696 (MCP server management panel), #719 (teams and agents management panel), #721 (Hermes Monitor dashboard reference), #722 (Aegis and App Data tab integration)

Assessmentadvisory
feature●●● hard80% confidence

Architectural feature requiring a new MCP server implementation and push-based data flow across all dashboard panels.

Likely files
  • api/mcp/server.py
  • api/routes/dashboard.py
  • static/js/app.js
  • static/js/panels/agent_status.js
  • static/js/panels/cost_tracking.js
  • api/gateway/mcp_gateway.py
  • server.py
Create the request

This opens a fresh agent run and a draft PR for issue #733.

Cancel